Feature extraction and integration for the quantification of PMFL data

Abstract

If the vast networks of aging iron and steel, oil, gas and water pipelines are to be kept in operation, efficient and accurate pipeline inspection techniques are needed. Magnetic flux leakage (MFL) systems are widely used for ferromagnetic pipeline inspection and although MFL offers reasonable defect detection capabilities, characterisation of defects can be problematic and time consuming. The newly developed pulsed magnetic flux leakage (PMFL) system offers an inspection technique which equals the defect detection capabilities of traditional MFL, but also provides an opportunity to automatically extract defect characterisation information through analysis of the transient sections of the measured signals. In this paper internal and external defects in rolled steel water pipes are examined using PMFL, and feature extraction and integration techniques are explored to both provide defect depth information and to discriminate between internal and external defects. Feature combinations are recommended for defect characterisation and the paper concludes that PMFL can provide enhanced defect characterisation capabilities for flux leakage based inspection systems using feature extraction and integration.